pcplantdb

[Chad Knepp <pyg@galatea.org>]

Richard Archer writes:
 > At 1:44 PM -0500 25/8/04, Chad Knepp wrote:
 > 
 > >Name, Culture, and Uses (and possibly Identification).  Remember that
 > >relationships with other things is information contained outside the
 > >plant itself.  In this light Uses becomes a bit fuzzy because a lot of
 > >uses are really relationships with other elements.  Anyway, lets do
 > >this again.  Maybe a good way to think about this is, if you wanted to
 > >know everything important about a plant what would that consist of.
 > 
 > If we're eventually going to be using this database to identify
 > potential guilds, then the more information stored about each plant
 > the better. And when I say "guild" I use the term very loosely.
 > What I really mean is the ability to use this database to find
 > a group of plants (or things) that don't have conflicting demands
 > and in which each element supplies some of the needs of other
 > elements in the group.
 > 
 > Finding groups of things (including plants, animals, insects,
 > structures and even processes) which work together in this way
 > is a fundamental aspect of permaculture, so any information we
 > can store which helps in this task would be useful.
 > 
 > 
 > Which leads to:
 > 
 > >Since it seems like a hot topic, I have some questions about
 > >relationships as well.  Are there a finite number of
 > >relationships?... for example would 1000 different kinds of
 > >relationships between elements adequately describe things... 100, 500,
 > >2000?  Would these relationships be one word, several words, a
 > >paragraph, or what?  Also if their are a finite number of Relationsips
 > >is a relationship in our system adequately expressed as "(one or more
 > >elements) has/have 'Relationship' with (one or more elements)"?
 > 
 > The concept of external relationships strikes me as being pretty open
 > ended. For example, every plant which can be a nitrogen fixer has an
 > implied relationship with every plant which is a nitrogen consumer.
 > It would not be desirable to require that every such potential
 > relationship be explicitly defined in the "relationships" table in
 > order that this relationship be used to identify a guild.

There are a number of implicit relationships that could be defined
explicitly.  So far in PFAF/Eden I just make this connection in my
head, like, 'oh it's a nitrogen fixer', or 'oh it's an aquatic plant'
and I know what that implies.  I'm not sure how much we want to point
out things... of course what is obvious to me/us may not be so to
others.

 > Instead the attributes of the objects (i.e. nitrogen fixing vs
 > nitrogen consuming) can be used to identify this relationship.

Yeah, nitrogen is tricky because it could be defined as an attribute
or a relationship.  At this point I think it would be best to have it
as an attribute and imply relationships from it.

 > So it follows that I think it would be best to expand the list of
 > attributes stored about each object so that these relationships can
 > be determined procedurally from the data set.

Although this is moving into implementation details I'm planning on
allowing methods to define element sets, so it would really be pretty
easy to say (every nitrogen fixing plant) has a relationship of
'providing nitrogen' to (all plants) or some such... although this
isn't the best example.  Better one would be (every Brassica) or
(every Tree with Dappled Foilage), etc...

 > Perhaps storing a description of relationships between specific
 > objects would be a useful thing in that it allows the entire data
 > set to be scanned to find other similar relationships which occur
 > between other objects.

One of the reasons I was asking about defining relationships is
because limiting them to a set makes them more searchable.  Another
idea I've had is to allow as much detail as desired to define the
relationship but also give it some broader classification for
scanning. 

 > But at the end of the day I think the end user will need to be able
 > to define how they want to analyse the data to find relationships.
 > And then it's going to be the quality and breadth of information
 > stored about each object that will really make this database useful.

Yup... so what are we missing?

 > >The thing about references is that most of our authorship from this
 > >point forth will not be accompanied references other than an allusion
 > >to personal experience.  I also do not find a reference to a book I do
 > >not own to be particularly helpful in anyway.  Online references could
 > >be helpful in the way you suggest, but I think that peer reputation
 > >will be the most widely used mechanism of ensuring authority.  I'm ok
 > >including references, but I suspect they will be unused.  What do
 > >other folks think?
 > 
 > This raises an interesting point, which Lawrence with his experience
 > with wikis might be able to shed some light on.
 > 
 > What happens if someone enters some data, for example plant X grows to
 > height 10 metres in locality A. Someone else then comes along and
 > enters that plant X grows to 15 metres in locality B.

This is the (LS) Locale Specific flags that I was putting in the
attributes.  Planting and fruiting characteristics vary every 100
miles or so.  A locale is another object type we need to think
about...  Locale specific information is essentially unique to users
from that local.

 > Does the initial data get overwritten? Is there some meaningful way
 > of storing and using both data points? Do we just look at which author
 > has the highest reputation and discard the other author's data?
 > If the second author's data gets discarded then how does their reputation
 > ever increase?

Oh boy do I have a plan!  Data will not generally be discarded.  Our
moderation system will allow virtually every piece of information to
be marked as helpful/correct or unhelpful/incorrect.  Just like an
internet search engine the information rated as most helpful/correct
will float to the top.  Information that is marked as helpful/correct
will increase the authors reputation.  Having a high reputation has
two advantages: 1) Information they post will automatically be given a
higher rank than lower rated authors.  2) Authors with a high
reputation will have a bigger impact on the rating of information they
moderate.  There are several scenarios where information can be
deleted.  The one I like best is that when a piece of information
receives only negative moderation from 2-4 unique authors (the number
possibly dependent on reputation) then it is deleted.  The challenging
aspect of this will be displaying all of this information in a
non-confusing way.

-- 
Chad Knepp
python -c 'import base64;print base64.decodestring("cHlnQGdhbGF0ZWEub3Jn")'